Deterministic and stochastic allele specific gene expression in single mouse blastomeres

Abstract

Stochastic and deterministic allele specific gene expression (ASE) might influence single cell phenotype, but the extent and nature of the phenomenon at the onset of early mouse development is unknown. Here we performed single cell RNA-Seq analysis of single blastomeres of mouse embryos, which revealed significant changes in the transcriptome. Importantly, over half of the transcripts with detectable genetic polymorphisms exhibit ASE, most notably, individual blastomeres from the same two-cell embryo show similar pattern of ASE. However, about 6% of them exhibit stochastic expression, indicated by altered expression ratio between the two alleles. Thus, we demonstrate that ASE is both deterministic and stochastic in early blastomeres. Furthermore, we also found that 1,718 genes express two isoforms with different lengths of 3'UTRs, with the shorter one on average 5-6 times more abundant in early blastomeres compared to the transcripts in epiblast cells, suggesting that microRNA mediated regulation of gene expression acquires increasing importance as development progresses.

Figure 1. Principal component analysis (PCA) of single cells of preimplantation embryos.

Two mature oocytes, eight two-cell blastomeres, six four-cell blastomeres, six eight-cell blastomeres, two trophectoderm cells, nine ICM cells, three epiblast cells are independently clustered.

Figure 2. Schematic diagram of allele specific gene expression (ASE) in individual blastomeres of a two-cell embryo identified by SNPs in mRNAs.

The two alleles are either expressed equally (balanced as in blastomere #1), or differentially (as in blastomere #2). Allelic ratio = counts of allele A/counts of allele a. If the allelic ratio of the gene is different in the two nearly identical blastomeres from the same two-cell embryo, it is denoted as AI. AI =  (allelic ratio in blastomere #1)/(allelic ratio in blastomere #2). The RNA-Seq reads across the SNP (here using nucleotide ‘A’ and ‘C’ as examples) can discriminate between the two alleles, and reveal relative abundance of expression. Sequence polymorphism is indicated (*).

Figure 3. The allelic ratios based on single cell RNA-Seq.

Allelic ratios of two mature oocytes (A) obtained from two adult female mice with different genetic background show all 9 possible clusters corresponding to the 9 genotypic combinations between the two cells. Allelic ratios of two mature oocytes obtained from the same adult female mouse (B) show 3 possible genotypes between the two cells: homozygous reference (AA/AA) and alternative (aa/aa) allele, and heterozygous (Aa/Aa). Here ‘A’ means reference allele, whereas ‘a’ means alternative allele. Allelic ratios of twelve nearly homozygous ES cells (C) show the homozygous expression of essentially all (>99.7%) expressed genes. Allelic ratios of two blastomeres of a two-cell stage embryo (D), a four-cell stage embryo (E), and an eight-cell stage embryo (F) show both homozygous and heterozygous loci. To avoid singularity for homozygous loci, we added 1 to the counts when we calculated the allelic ratios.

Figure 4. Determination of ASE and AI.

Homozygous distribution of the minor allele frequency of 12 mouse ES cells obtained from the same ES cell line, generated from nearly homozygous embryo (blue curve). The distribution of the allele ratio shows that there is a 95% chance for a homozygous locus with at least 25 counts to generate a minor allele frequency of 0.066 or lower. Distribution of minor allele frequency for one blastomere in a two-cell embryo is shown in green. The balanced allele distribution of minor allele frequency for one blastomere from a two-cell embryo is shown in red. The balanced heterozygous expression (equal expression of both alleles) is expected to have their allele frequency of 50%. The distribution generated from all heterozygous loci of an individual cell represents an approximation of the distribution of allelic ratios of balanced loci, that takes in consideration library and instrument biases. Locus is called as having allele specific expression (ASE) if the resulting p-value is <0.01, which corresponds to minor allele frequency of up to 33%, and major allele frequency of at least 67%. The distribution of two-cell blastomere with minor allele frequency above 50% is due to smoothing function produced by R scripts.

Figure 5. Schematic diagrams (A) and coverage plots (B) that express short and long 3′UTR isoforms.

Co-expression of two types of mRNAs of long (distal) and short (proximal) 3′UTR from the same gene, such as Cbx5, Tet1, and Gsk3b. Here use Sox2 as a reference, which has only one short 3′UTR. The 3′ bias of single cell RNA-Seq method can discriminate between the long and short 3′UTRs of the mRNAs from individual genes by looking at the shape and the spread of the cover plot of an expressed gene.

Figure 6. Features of alternative 3′UTRs during preimplantation development.

Density plots of CPSF binding motif distance from predicted 3′UTR site. For all cell stages analyzed, CPSF binding motifs are observed 20 nt upstream of the predicted cleavage site (A) as expected, in concordance with CPSF location relative to known RefSeq cleavage sites (B). (C) The ratio of expression levels between the proximal (short) and distal (long) 3′UTR of the same genes during preimplantation development.

Figure 7. Potential regulation of long and short 3′UTRs of the same gene within individual cells by microRNAs.